I'm not an economist, but will play one on the net for a few minutes.
I'm not too concerned about the overhead for transaction costs; all that
will do is set a floor on the smallest transaction. If someone can make
money charging $.001 per transation, they'll be fat, dumb and happy.
If they lose money, they'll go out of business -- and the next-lowest
price provider will take over. Somewhere in that process, we'll get
overhead charges for transactions which are just slightly less than the
cost of providing the transaction.
Fraud will factor in. If efraud is trivially easy on unsecured
transations, then the overhead cost savings obtained by not securing the
transaction will be eaten up by the cost of fraud, and no-one will offer
unsecured transations. If securing the transaction is hideously expensive
(I doubt it); then only large transations will be done -- because only
with large transations will the overhead cost be `lost in the noise.'
But somewhere a balance will come, and it'll be dictated by overall
economic costs rather than simply CPU speed or protocol overhead.
Damned near *anything* you do with a computer is cheaper than having
human beings handle paper, and IMHO that includes a few tenths of a
second to do encrypt/decrypt. If a service is so large and doing so many
transactions per second as to require a Cray, then you can bet they will
charge enough overhead to pay for that Cray. Or for special ASICs, or
roomsful of Alphas, or whatevers. Remember, we don't need perfect
security or perfect encryption -- just enough to make fraud more costly
than the profit. Given what I've seen of the market and technology,
that doesn't seem a problem to me.
But we're getting pretty far afield of firewalls here, so I'll shut
``A Woollcott second edition.''
-- Franklin Peirce Adams, replying to Alexander Woolcott's boast:
`What is so rare as a Woollcott first edition?'